A method of online anomaly perception and failure prediction for high-speed automatic train protection system

https://doi.org/10.1016/j.ress.2022.108699Get rights and content

Highlights

  • The temporal correlation of ATP operation logs is revealed based on LSTM network.

  • Online anomaly perception of high-speed automatic train protection system.

  • A fault prediction method of ATP system based on time series is proposed.

  • There exists high-dimensional dependence between the evolution behavior of ATP failure rate time series and the historical data.

  • ATP intelligent operation and maintenance data service platform is designed.

Abstract

Automatic train protection (ATP) system is the key to ensure the safe operation of high-speed trains. However, the existing operation and maintenance mode for ATP systems cannot diagnose fault in time. In order to improve the protection capability of trains, this paper proposes an online anomaly perception and failure prediction method. First, with real-time operating data, an anomaly perception model based on long short-term memory network is established, where unstructured data are parsed into structured log keys and parameter vectors. It is trained with sequence matrices and its learning performance under different parameters is tested to find the optimal model. Experimental results show that the classification accuracy is 0.981, which is better than the existing methods. Then, with historical data, a failure prediction model based on time series is established, where one-dimensional time series of failure rate are reconstructed to high-dimensional space. The support vector regression method is used to fit the complex functional relationship between phase point and predicted point. And different algorithms are taken to find the optimal parameters. The results show that the model has the strongest generalization ability with the accuracy of 0.987. Finally, the intelligent operation and maintenance data service platform is designed.

Introduction

By the end of 2021, the operating mileage of China's high-speed railway had been more than 40,000 kms, exceeding half of the world's mileage. China has become the country with the longest high-speed railway mileage, the highest transportation density and the most complex operation scenario in the world. Automatic train protection (ATP) system is the key to ensure safe and efficient operation of the train and is known as the nerve center. In 2020 and 2021, under the impact of COVID-19, although some trains stopped running to reduce the population flow, ATP systems still ran 3120.2 million kilometers average per day. By calculating the speed profile in real time, it automatically adjusts the space distance with the front train and controls operation position and speed of current train. It is installed at both ends of the high-speed train and adopts redundant structure to connect with other external equipment such as train and monitoring system. It mainly consists of vital computer (VC), speed & distance processing unit (SDU), balise transmission module (BTM), track circuit reader (TCR), driver machine interface (DMI), train interface unit (TIU), GSM railway (GSM-R), juridical recorder unit (JRU), etc. The sub equipment cooperates with each other to control the train, and any failure will affect the normal operation of ATP systems.

There are five types of ATP systems: 300T, 300S, 300H, 200H and 200C, and each has several failure modes. For instance, BTM of 300T-type related failures are divided into run-time balise service available (BSA) anomaly, start-up BSA anomaly, BTM status telegram invalid, BTM unable to correctly parse messages, routine tests anomaly, etc. Once the system fails, accurate fault diagnosis is difficult. Therefore, the efficient operation and maintenance of ATP systems faces great challenges.

At present, during the operation and maintenance of ATP systems, whole process monitoring and regular maintenance are adopted to maintain equipment reliability and prevent major accidents. However, since the operating conditions, external environment and sudden factors vary greatly, monitoring mechanism cannot accurately diagnose anomalies and regular maintenance cannot effectively maintain the health of equipment. This strategy faces the following challenges.

(1) The challenge of safe operation. The currently adopted operation and maintenance method based on post-processing [1] has disadvantages such as difficulty in fault diagnosis and low efficiency of failure handling. This passive mode cannot meet the requirement of fast and accurate fault location and emergency response, resulting in a large scope of the fault. For example, the ATP system fault affect 784 trains nationwide in 2021, with a delay of 7286 min.

(2) The challenge of intelligence. System fault diagnosis, maintenance implementation, operation and maintenance resource configuration, etc., are almost completed manually, and the level of intelligent perception is low, resulting in heavy tasks and insufficient operation capabilities. What's more, operation data are not effectively mined.

(3) The challenge of human resources. The operation and maintenance workload of ATP systems is heavy, the work content is repetitive and boring, and the attractiveness of operation and maintenance positions, especially on-duty positions is declining. The contradiction between operation and maintenance requirements and human resources has become inevitable in the development of ATP systems and even high-speed railways.

Intelligent operation and maintenance refers to the method that uses artificial intelligence algorithms such as machine learning to automatically learn and summarize rules from massive amounts of data, so as to make decisions. In order to solve the above problems, based on intelligence, this paper forms a comprehensive operation and maintenance security system that integrates anomaly state perception and failure prediction. It also forms a production organization model for intelligent operation and maintenance of high-speed railway equipment and realizes the rapid and accurate anomaly detection, accident prevention and emergency response, and dynamic configuration of transportation resources, so as to improve the protection capability and operation efficiency of high-speed trains. The main contributions of this paper are as follows.

An online anomaly perception model of ATP system based on long short-term memory network is proposed, which can realize online anomaly detection for log keys and parameter vectors, and solve the problem of delay in fault diagnosis and response, and inaccurate fault location.

A fault prediction method of ATP system based on time series is proposed, and the high-dimensional dependency between the future trend of failure rate and historical data is revealed. It helps to predict the evolution trend of the future state, and realize the refined and condition-based maintenance of the whole life cycle.

The ATP system intelligent operation service platform is designed, and the architecture of ATP system combined with the vehicle and the ground information is built. Finally, the refined operation and maintenance of the whole life cycle of each ATP system is realized.

The rest of this paper is organized as follows. In Section 2, we review the related work. In Section 3, an anomaly perception model of ATP systems is established. In Section 4, the failure prediction model is built. In Section 5, experiment and analysis are done. In Section 6, intelligent operation and maintenance system is designed. In Section 7, the conclusion of our work is drawn.

Section snippets

Related work

With the deployment of industry 4.0 strategy, intelligence had become an important direction for development of global railways [2]. Intelligent high-speed train had become a new competition hotspot in the world. It was pointed out that maintenance costs accounted for 15% to 60% of the total costs of all manufacturing industries [3]. As an important part of intelligent high-speed train application and safety guarantee system, intelligent operation and maintenance had also become a hot spot in

Anomaly perception

Anomaly perception is a key step in building a safe and reliable system. System log is to record the state and major events during the operation, which is helpful for fault detection and root cause analysis. It is important resource for understanding the system state and performance problems. Therefore, it is good source for online detection and anomaly state awareness.

The real-time log generated during the operation of the ATP system can be regarded as a sequence element that follows certain

Failure prediction

In Section 3, the present state anomaly perception is realized. In this section, the future failure prediction is achieved. The two critical technical difficulties of intelligent operation and maintenance are solved from time dimension.

Experiment and analysis

In order to verify the effectiveness and accuracy of the anomaly perception and failure prediction model, the real operation and maintenance data of a certain type of ATP systems are selected as samples. Then the model performance under different parameter combinations is compared to find the optimal model for actual scenarios.

Discussion and application

The above sections describe the key technologies for intelligent operation and maintenance of ATP systems from the time dimension. Operating state perception is for the present and failure prediction is for the future. Based on these, an effective estimation of the present and future state of the system is achieved, and automated tools are used to implement operation and maintenance decisions. Therefore, a new intelligent operation and maintenance architecture has been created for ATP systems.

Conclusion

For the present, an anomaly perception method based on LSTM neural network is proposed for high-speed automatic train protection system, where the anomaly detection of log key and parameter vector is realized online. By randomly selecting the operation data of 300T-type ATP systems for experiments, the validity of the model is verified, and the accuracy is better than the existing anomaly detection methods. In addition, the model has strong stability with parameter changes, which is suitable

CRediT authorship contribution statement

Renwei Kang: Conceptualization, Methodology, Software, Writing – original draft, Formal analysis. Junfeng Wang: Conceptualization, Writing – review & editing, Funding acquisition. Jianqiu Chen: Writing – original draft, Data curation, Funding acquisition. Jingjing Zhou: Software, Formal analysis, Investigation. Yanzhi Pang: Validation, Resources. Longlong Guo: Software, Formal analysis. Jianfeng Cheng: Investigation, Project administration.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgment

This work was supported in part by State key laboratory of rail traffic control and safety (under grant RCS2022ZT010), in part by Nanning youth science and technology innovation and entrepreneurship talent cultivation project (under grant RC20180109), in part by Guangxi science and technology plan project-innovation driven development project (under grant GUIKE AA21077011). The authors would like to thank Junli Li, she revised grammatical and syntactical errors in the manuscript.

References (42)

  • H. Nguyen et al.

    Deep learning methods in transportation domain: a review

    IET Intell Transp Syst

    (2018)
  • E.T. Bekar et al.

    An intelligent approach for data pre-processing and analysis in predictive maintenance with an industrial case study

    Adv Mech Eng

    (2020)
  • Y. Shi et al.

    Condition-based maintenance optimization for multi-component systems subject to a system reliability requirement

    Reliab Eng Syst Saf

    (2020)
  • K. Antosz et al.

    The use of artificial intelligence methods to assess the effectiveness of lean maintenance concept implementation in manufacturing enterprises

    Appl Sci Basel

    (2020)
  • X.Y. Xu et al.

    A novel vision measurement system for health monitoring of tunnel structures

    Mech Adv Mater Struct

    (2020)
  • J.R. Ruiz-Sarmiento et al.

    A predictive model for the maintenance of industrial machinery in the context of industry 4.0

    Eng Appl Artif Intell

    (2020)
  • A. Sharma et al.

    Artificial intelligence-based fault diagnosis for condition monitoring of electric motors

    Int J Pattern Recognit Artif Intell

    (2020)
  • J. Singh et al.

    A systematic review of machine learning algorithms for prognostics and health management of rolling element bearings: fundamentals, concepts and applications

    Meas Sci Technol

    (2021)
  • P. Wen et al.

    A generalized remaining useful life prediction method for complex systems based on composite health indicator

    Reliab Eng Syst Saf

    (2021)
  • C. Chen et al.

    A risk-averse remaining useful life estimation for predictive maintenance

    IEEE/CAA J Autom Sin

    (2021)
  • Z. Chen et al.

    Machine remaining useful life prediction via an attention-based deep learning approach

    IEEE Trans Ind Electron

    (2021)
  • Cited by (12)

    • Overview of Development Trends of Intelligent High-speed Rail Train Control Systems

      2024, Tiedao Xuebao/Journal of the China Railway Society
    View all citing articles on Scopus
    View full text